Abstract
We present a new automatic method to detect cones in AO retinal images. Images undergo topographical feature extraction, and the features are classified as either cone or not-cone. Out of five different binary classifiers tested, the one based on logistic regression achieved the best detection performance. A set of truth data consisting of manually identified cones was used to train the classifiers. Cone detection performance was compared quantitatively against two currently existing methods, and the proposed approach achieved the lowest error rate.
© 2014 Optical Society of America
PDF ArticleMore Like This
Shahab Chitchian, Adam R. Boretsky, Frederik J. van Kuijk, and Massoud Motamedi
BSu3A.17 Biomedical Optics (BIOMED) 2012
Michael Pircher, Franz Felberer, Julia S. Kroisamer, U. Schmidt-Erfurth, Bernhard. Baumann, Stefan Zotter, and Christoph K. Hitzenberger
FW5F.2 Frontiers in Optics (FiO) 2013
Radim Kolar, Ivana Liberdova, Jan Odstrcilik, Michal Hracho, and Ralf P. Tornow
104130A European Conference on Biomedical Optics (ECBO) 2017